Evolutionary algorithm for a genetic robot’s personality based on the Myers–Briggs Type Indicator

https://doi.org/10.1016/j.robot.2012.01.007Get rights and content

Abstract

The genetic robot has many configurable genes that contribute to defining the robot’s personality. The large number of genes allows for a highly complex system, however it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the robot’s personality while manually initializing values for the individual genes. To overcome this difficulty, this paper proposes MBTI-EAGRP. It is a fully autonomic gene-generative algorithm for a genetic robot’s personality in a mobile phone. After grasping the user preferences through MBTI assessment using the neural network algorithm, the evolutionary algorithm generates and evolves a gene pool that customizes the robot’s genome so that it closely matches a simplified set of personality features preferred by the user. Finally, an evaluation procedure for individuals is carried out in a virtual environment using tailored perception scenarios and real MBTI measurements.

Highlights

► An MBTI-based-Evolutionary Algorithm for a Genetic Robot’s Personality is proposed. ► The Self-Organizing feature Map establishes a user preference through the MBTI. ► MBTI-EAGRP characterizes a variety of internal states and their concomitant behaviors. ► MBTI-EAGRP is demonstrated by a software robot in a mobile phone and a real robot.

Introduction

This paper defines the genetic robot as a robot which has its own genetic code and so aims to autonomically generate the genetic code that can reliably reflect the emotional personality which a user prefers. To achieve this, a user has only to measure his/her psychological preferences through the Myers–Briggs Type Indicator (MBTI) assessment. After grasping the user types, the MBTI-based evolutionary algorithm generates and evolves a gene pool that customizes the genetic code so that it closely matches a simplified set of preferred personality features.

Gene on a robot is a very powerful entity with its implicit goals and the way it functions [1], [2]. As we and other animals are machines created by our genes [3], [4], genetic encoding and algorithms for modularity and reusability can serve economically as an engine for consistency and coherence [1]. Namely, the genetic code has the main functions of reproduction or reusability, and evolution. Thus the genes are considered as key components in defining a creature’s personality and the essence of this research should be on the genomes of various types of artificial creatures, such as pet-type, humanoid-type, or head-type, which can be implemented in either a real hardware robot or a simulated software robot (sobot) [5], [6], [7], [8], [9], [10].

Personality is the engine of behavior [11]. Considering the user interactions, its personality is crucial in building a believable emotional robot. Having a diverse personality is important because it can be encoded as an inherited trait, which decides the behavior based on an internal state in response to the stimulus. A trait is characterized by the Big Five personality dimensions [12], [13]. This allows for the creation of diverse personalities for the agent, e.g., allowing it to express highly extroverted and at the other end of the scale, highly introverted characteristics.

The large number of genes allows for a highly complex system [14]. However it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the robot’s personality while manually initializing and tuning values for the individual genes [11], [15]. To overcome this difficulty, MBTI can grasp the personality types of genetic robot users [16] and provide us with the standard to which types of genetic robots should be made: one is to make a genetic robot that has similar personal characteristics to the user (related to dominant and auxiliary processes in the MBTI result), the other is to make a genetic robot that has the opposite personal characteristics to the user (related to the third and inferior processes).

An MBTI-based-Evolutionary Algorithm for a Genetic Robot’s Personality (MBTI-EAGRP) is proposed. MBTI-EAGRP is a novel algorithm to identify the generative mechanism that characterizes a variety of internal states and their concomitant behaviors by simply measuring a user preference through the MBTI assessment.

MBTI-EAGRP consists of two phases to generates the robot genome. The first phase, SOMMBTI, is to measure a user feature through the MBTI assessment, to make the user feature map by the self-organizing feature map (SOM) using the neural network algorithm, and to establish the required robot personality preference based on SOM. The second phase, EAGRP, is to evolve a gene pool that customizes the robot’s genome so that it closely matches a simplified set of preferences generated by the first phase. The evaluation procedure for individuals is carried out in a virtual environment using tailored perception scenarios. The genetic robot is validated by implanting the robot genome generated by MBTI-EAGRP into a virtual creature, AnyRobot in a mobile phone.

The remaining structure of this paper is outlined as follows: Section 2 briefly introduces an artificial creature, AnyRobot, as a genetic robot and its internal architecture. Section 3 proposes SOMMBTI to establish the user preference gain. Section 4 describes the structure of EAGRP from a theoretical viewpoint. Experiments are carried out to demonstrate the performance of MBTI-EAGRP in Section 5. Then empirical and theoretical analyses follow to investigate its characteristics. Concluding remarks follow in Section 6.

Section snippets

Genetic robot

This section introduces a software robot, AnyRobot [17], [18], [19], [20], and the novel concepts of the artificial chromosome, the robot genome and the genetic robot are proposed through AnyRobot.

Self-organizing feature map for the MBTI (SOMMBTI)

The Myers–Briggs Type Indicator (MBTI) assessment [21] is briefly summarized in Section 3.1. The 3-D SOM of the user’s feature through the MBTI is represented and expanded from the original version of the 2-D SOM by using self-organizing feature map (SOM) (refer to Section 3.2). In consequence, Section 3.3 proposes a method to establish the required personality preference based on SOM. After grasping the user preferences, two types of genetic robot’s personality can be made, having similar or

MBTI-based evolutionary algorithm for a genetic robot’s personality (MBTI-EAGRP)

MBTI-EAGRP uses a novel representation of the robot genome for the individual that was introduced in the previous Section 2.2. The algorithm uses its own evolutionary techniques for generating a desired genetic robot’s personality.

Experiments

This section defines the parameters for the genetic robot and employs the MBTI results that a user obtained, as shown in Fig. 5. The experimental results verify the feasibility of MBTI-EAGRP.

Concluding remarks

This paper devised a genetic robot with a personality and its genes. The MBTI-EAGRP was proposed as its gene-generative algorithm using the MBTI assessment, the neural network algorithm, and the evolutionary algorithm. The most difficult step in this procedure was in generating an artificial life form with a personality that was both complex and feature-rich, but still plausible by human standards for an emotional life form. This was demonstrably achieved via the MBTI-EAGRP mechanism and its

Acknowledgments

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (grant number: 2011-0015064).

Kang-Hee Lee received B.S., M.S., and Ph.D.degrees in electrical engineering and computer science from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1999, 2001, and 2006, respectively. Since 2006, he has been a Senior Engineer in Digital Media & Communication Research Center, Samsung Electronics Company, Ltd., Korea. Since moving to Soongsil University in 2009, he is currently an Assistant Professor in the Global School of Media, Soongsil University, Seoul,

References (25)

  • R. Arkin et al.

    An ethological and emotional basis for human–robot interaction

    Robotics and Autonomous Systems

    (2003)
  • A. Sloman

    Designing emotions for activity selection in autonomous agents

  • E. Roll

    A theory of emotion, its functions, and its adaptive value

  • R. Dawkins

    The Selfish Gene

    (1976)
  • R. Dawkins

    The Blind Watchmaker

    (1986)
  • B.M. Blumberg, Old tricks, new dog: ethology and interactive creatures, Ph.D. Dissertation, MIT, Cambridge, MA,...
  • H. Miwa, T. Umetsu, A. Takanishi, H. Takanobu, Robot personality based on the equation of emotion defined in the 3D...
  • C. Breazeal

    Designing Sociable Robots

    (2002)
  • C. Elliott et al.

    Autonomous agents as synthetic characters

    AI Magazine

    (1998)
  • J.-H. Kim et al.

    The origin of artificial species: genetic robot

    International Journal of Control, Automation and Systems

    (2005)
  • A. Ortony

    On making believable emotional agents believable

  • P.T. Costa et al.

    The NEO Personality Inventory Manual

    (1985)
  • Kang-Hee Lee received B.S., M.S., and Ph.D.degrees in electrical engineering and computer science from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1999, 2001, and 2006, respectively. Since 2006, he has been a Senior Engineer in Digital Media & Communication Research Center, Samsung Electronics Company, Ltd., Korea. Since moving to Soongsil University in 2009, he is currently an Assistant Professor in the Global School of Media, Soongsil University, Seoul, Republic of Korea. His current research interests include the areas of ubiquitous robotics, evolutionary robotics, emotional robotics, educational robotics, and media robotics.

    Younggeun Choi received a B.S. degree in aerospace engineering and electrical engineering in 1998 from the Korea Advanced Institute of Science and Technology (KAIST), Daejon, Korea, a M.S. degree in computer science in 2005 and a Ph.D. degree in computer science in 2010 from the University of Southern California, Los Angeles. He is currently an Assistant Professor in the Department of Applied Computer Engineering, Dankook University, Republic of Korea. His current research interests include the design and control of rehabilitation robots, haptics, neuroscience, and machine-learning-based modeling.

    Daniel J. Stonier has received degrees in electrical engineering and mathematics from the University of Queensland, Australia. He also went on to complete a Ph.D. in mathematics in 2002. His previous work history began with teaching in mathematics, before switching to engineering when he moved to Korea for a post-doctoral period at KAIST. Looking to obtain more practical experience, he moved to Yujin Robot, where he has since worked as a developer and lead for the control group. His current research interest is in the field of monocular based slam, however the scope of work at Yujin is extremely varied and subsequently other interests closely align with the field of endeavor there. This includes slam, arm manipulation, motor control, 3-D sensing and software frameworks at a higher level for simplifying the robotic development process (working in and around RoS).

    View full text